2019 TIBCO Analytics Forum - Hackathon

Last updated:
4:36pm Nov 14, 2019

TIBCO Analytics Forum Hackathon


Congrats to the 2019 Winners!!!

Top finishers shown above with TIBCO Chief Analytics Officer, Michael O'Connell, and Data Scientists, Neil Kanungo and Andrew Berridge.

  • 1st - Ivan Leung, Oxy

  • 2nd - Derrick Roberts, Enterprise Products

  • 3rd - Atheer Al-Attar, Enterprise Products

  • 4th - Richard Barthel, Texas Instruments


Overview

At TIBCO we believe in learning by doing, and being an innovator that pushes the boundaries of the ordinary. In keeping that spirit, we are proud to host our Hackathons which encourage our users to test their knowledge and try new things. This year, almost 50 data guru's competed in our Hackathon for the ultimate prize! It was a packed room with lots of head-scratching, eureka moments, and overall analytics fun!

See also highlights from the 2016 Hackathon and 2017 Hackathon on our Community.

Challenge Details

This year there were 3 challenges covering the following topic areas:

  1. Geoanalytics (30 min)
  2. Multivariate Analysis (40 min)
  3. Streaming (40 min)

The challenges involved oil & gas well production and drilling data to be used with TIBCO Spotfire and TIBCO Streaming. All software was provided on an Amazon Web Service AMI so participants simply had to login to the environment and each were guaranteed to have identical hardware and software capabilities. To try the challenges yourself, download the Hackathon DXP attached to the bottom of this page!

Rules

Click here for the official 2019 Hackathon rules.

Solution Overview

>> CHALLENGE 1: GEOANALYTICS

Participants must take the provided well production data by Texas county, and convert units for oil and gas into Barrel of Oil Equivalents (BOE). This was typically done through calculated columns though inserting column transformations was also a valid approach. After calculating the BOE, each county should have been visualized by county polygon boundaries, and colored by Total BOE:

The county locations were automatically added through automatic geocoding in Spotfire, however, participants must have chosen to represent this geocoded data as a "Feature Layer". After visualizing Total BOE, pie chart markers were required to subdivide each county by the production type.

>> CHALLENGE 2: MULTIVARIATE ANALYSIS

This was perhaps the most difficult challenge of the three. This challenge required users to access external data via a REST API service, join that with existing embedded data, and then perform a series of multivariate analyses. To access the REST API data, the participant simply had to copy/paste the provided script into a data function, and include the start/end dates of 1/1/2018 and 1/1/2019. They then needed to find the variables with R2 > 0.3 across the production data when compared to OilVol. This is easily done by using the built-in Data Relationships tool in Spotfire to compare OilVol with all other numerical variables:

The results of this show the production values for oilRate, gasRate, gasVol, waterRate, and waterVol all having Rsq > 0.3. To compare profile all wells by these variables and find the 5 most similar wells, a line chart can be constructed as a parallel coordinate plot and then used with Spotfire Search and Line Similarity to drill in by multiple variables:

A Heirarchical Clustering method could also be used here with the same variables. Credit was given to any participant showing a reasonable method for determining similarity to the specificed Well ID. As a bonus, users may visualize the production across all wells using a heatmap on a map chart:

>> CHALLENGE 3: STREAMING

This challenge involved creating a data stream using the new Spotfire Streaming capabilities and then developing basic visualizations to monitor drilling performance. To create a streaming data set, users were invited to use the Streaming Connectivity Wizard with a provided CSV data set to create a "Feed Simulation." The first visualization required visualizing the Bit Depth vs Hole Depth across each Rig ID. The second visualization required visualizing the past 5 seconds of RPM data per Rig ID, which required participants to limit the data window to 5 seconds in the visualization properties:

For the bonus, particpants needed to create a Color Rule tied to a Document Property and Property Control slider to adjust colors of Torque above a specified threshold on the slider:

Spotfire Resource Links

  • TIBCO Answers page is a question and answer forum for TIBCO Products. Spotfire questions can be asked under Analytics or Spotfire tags.

  • Spotfire Enablement Hub is the link to Spotfire X training videos and resources.

  • Dr Spotfire hosts monthly 30-minute Feature Sessions for both new and existing Spotfire users. Subscribe to our YouTube channel to get weekly bite-sized Spotfire Tips.

  • TIBCO Community Exchange contains under category "Analytics" reusable components and data functions that can be downloaded and then added to your dxp file

 

 Back to TIBCO Analytics Forum Main Page